Workshop Engineering and Music

"Human Supervision and Control in Engineering and Music"

Toshiyuki Inagaki

Supervisory Control in Music and Engineering: Dissimilarities and Their Implications

Abstract

This note discusses similarities and dissimilarities between two supervisory control frameworks, one for music and the other for engineering. Especially, this note tries to clarify differences in characteristics of a task-interactive computer (or its parallel).

1. Supervisory Control

When discussing supervisory control in engineering applications, we usually distinguish human operators, a human-interactive computer (HIC), task-interactive computers (TIC), and a controlled process. As Prof. Dr. Johannsen writes in his “Short Description” of this International Workshop, “The concept of Supervisory Control can be applied to every orchestra under a conductor as Supervisor.” The concertmaster and part leaders serve as HICs, remaining individual players correspond to TICs, and a musical piece may be regarded as a controlled process, though human players are not computers.

It is almost straightforward to rephrase the conductor’s role in terms of the conventional supervisory control notions, (1) planning, (2) teaching, (3) monitoring, (4) intervening, and (5) learning. It is also easy to imagine how significant roles the concertmaster and part leaders play in interpreting and transmitting the conductor’s ideas or intentions to players in the orchestra. (When this author was playing the violin at a university orchestra in his student days, he used to be helped greatly by the concertmaster and the leader of his part when he failed to interpret a professional conductor’s sophisticated yet complicated gesture.) Once tasks are expressed clearly, each player does his/her best to accomplish the specified tasks with his/her sensory and control capabilities.

In addition to these functional or organizational similarities, we may see psychological similarities, such as trust/distrust/over-trust, reliance/over-reliance, between musical and engineering supervisory control frameworks. However, it may not be surprising to find dissimilarities between the two frameworks. Among them, this short note will discuss dissimilarities on TIC characteristics.

2. Interactions among TICs

The TICs in engineering systems do not communicate directly with each other. They perform their tasks independently based on directives given by the HIC. Each TIC is associated with sensors to measure values of parameters (or variables) that are necessary in implementing its feedback control.

However, TICs cannot be totally independent with each other even in engineering systems. If a controlled process is equipped with several TICs, a control input by a TIC may be regarded as “disturbance” by some other TIC in the system.

A typical example of this type of interaction between two TICs can be seen in the Airbus A300-600R crash at Nagoya in 1994. At some point during the final approach, the pilot flying gave a Go-Around directive to the automation (TIC) unintentionally. The TIC started its maneuver for going around. However the pilot initiated an action to descend for landing, where the pilot can also be seen as another TIC. Thus intentions between the two TICs became completely contradictory. For the automation, the human’s input to descend was a disturbance that must be cancelled out by applying stronger control input. From the viewpoint of the pilot, the aircraft did not descend smoothly and thus he applied stronger control input. With a lack of communication between the two TICs, the aircraft finally became unstable.

In supervisory control for music, individual players (TICs), although they are basically responsible for their own parts, may communicate with each other spontaneously. A TIC in a part may listen to what and how TICs in some other parts are playing. This kind of spontaneous communication may enrich orchestra performance.

The author had a chance to play the Tannhaeuser Overture by Richard Wagner in the aforementioned university orchestra. As you may know, the strings, especially the violins, are tasked incredibly. When the woodwinds and the brass enjoy playing the majestic motive with whole notes, the violins have to struggle against pages of “scales” with sixteenth notes. I, as one of TICs, felt no amusement when I was solely practicing the scales, even if I was successful in playing the long series of scales without any failure. The circumstance changed when playing in a full orchestra. I could catch how the woodwinds and the brass felt and wanted to play, and I could even enjoy incorporating their intentions into my playing the scales. Note that the communication was not made through the concertmaster, part leaders, or the conductor.

3. Cooperation among TICs in Engineering Systems

The Tannhaeuser Overture experience gives the author an idea that, if TICs could communicate with each other directly to share aims and intentions of control, they may be able to “coordinate” their controls nicely and appropriately. As has already been stated, TICs in engineering applications do not communicate with each other in a direct manner. It is supposed to be a task of HIC to coordinate actions of TICs, although not all HICs are performing the task in reality.

If TICs are given capabilities to communicate with each other, we may have to pay some costs. Potential increase in automation-induced surprise may be one of such costs. It would be valuable, however, to investigate possibility of the direct communication and cooperation among TICs, especially for systems in which no tyrannical HIC seems to exist, or in other words, for systems where each TIC has a post of HIC concurrently.